Analytical Chemistry Theoretical and Metrological Fundamentals K. Danzer Analytical Chemistry
Theoretical and Metrological Fundamentals
With 138 Figures and 31 Tables K. Danzer Am Friedensberg 4 07745 Jena Germany e-mail: [email protected]
Library of Congress Control Number 2006932265
ISBN-10 3-540-35988-5 Springer Berlin Heidelberg New York ISBN-13 978-3-540-35988-3 Springer Berlin Heidelberg New York DOI 10.1007/b103950
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in other ways, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under German Copyright Law.
Springer is a part of Springer Science+Business Media springer.com © Springer-Verlag Berlin Heidelberg 2007
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.
Product liability:The publishers cannot guarantee the accuracy of any information about dosage and appli- cation contained in this book. In every individual case the user must check such information by consulting the relevant literature.
Cover-Design: WMXDesign GmbH, Heidelberg Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Germany Production: LE-TEX Jelonek, Schmidt & V¨ockler GbR, Leipzig, Germany Printed on acid-free paper 2/3141/YL-543210 Preface
Analytical chemistry can look back on a long history. In the course of its de- velopment, analytical chemistry has contributed essentially to the progress of diverse ˇelds of science and technology. Signiˇcant progress in natu- ral philosophy, chemistry and other disciplines has been founded on re- sults and discoveries in analytical chemistry. Well known examples include Archimedes' principle, iatrochemistry, emergence and overcoming of phlo- giston theory, basic chemical laws on stoichiometry and mass proportions as well as the discoveries of elements and of nuclear ˇssion. As chemistry split into inorganic, organic and physical chemistry in the middle of the nineteenth century, the question occasionally arose as to whether analytical chemistry was likewise an autonomous ˇeld of chemistry. At the beginning of the twentieth century, analytical chemistry was estab- lished by Wilhelm Ostwald and other protagonists on the basis of chemical and physicochemical laws and principles. In addition, analytical procedures and instruments spread into other chemical ˇelds. However, in the 1950s and 1960s a new generation of analysts opened their mind to ideas coming from other scientiˇc ˇelds dealing with measurements from other points of view such as information science, metrology, technometrics, etc. As a result, efforts made towards the acceptance of analytical chem- istry as an autonomous ˇeld of chemistry increased. Analysts in USA and Europe founded organisations and forums that focussed ideas and devel- opments in progressive directions. In this way, not only were national and international societies formed but also regular conferences like Pittcon and Euroanalysis. In this connection the so-called Lindau circle should be men- tioned where analysts from Germany, Austria and Switzerland made public relevant fundamentals of analytical chemistry derived from statistics, infor- mation theory, system theory, signal theory, game theory, decision theory, metrology, etc. In parallel with this, such modern principles have been suc- cessfully applied in selected case studies, mainly by scientists from USA, UK, and Canada. From this development, real theoretical and metrolog- ical fundamentals of analytical chemistry have crystallized. However, the acceptance of analytical chemistry as an automomous branch of chemistry has not been realised. Gradually, however, progress in analytical chemistry culminated in the 1990s in the proclamation of a new scientiˇc discipline, analytical science. Protagonists of these activities included Michael Widmer and Klaus Doerffel. VI Preface Today, analytical chemistry is still a discipline within chemistry. Al- though characterized as an auxiliary science, analytical chemistry continues to develop and grow just as before. It is no detraction being characterized as an auxiliary discipline as mathematics shows. However, it is likely that efforts to make analytical chemistry a more independent science will be repeated in the future from time to time. Analytical chemistry possesses today a sound basis of chemical, physical, methodical, metrological, and theoretical fundamentals. The ˇrst of these are usually taken as the basis of classical textbooks on analytical chemistry. The others are found in diverse publications in the ˇeld of analytical chem- istry and chemometrics. It is essential to state that chemometrics is not the theoretical basis of analytical chemistry but it contributes signiˇcantly to it. Frequently, analytical chemistry is considered to be a measuring science in chemistry. Therefore, its object is the generation, evaluation, interpreta- tion, and validation of measuring signals as well as the characterization of their uncertainty. With this aim, the analyst needs knowledge of the general analytical process, statistics, optimization, calibration, chemometrical data analysis, and performance characteristics. In this book the attempt is made to summarize all the components that can be considered as building blocks of a theory of analytical chemistry. The \building" constructed in this way is a provisional one. It is incomplete and, therefore, extension, reconstruction and rebuilding have to be expected in the future. A large number of mathematical formulas will be found in the book. This may be regarded as a disadvantage particularly because some of them are not readily to apply in daily analytical practice. However, great scientists, explicitly Emanuel Kant, said that a scientiˇc branch contains only so much of science as it applies mathematics. Consequently, all the relationships which can be described mathematically should be so described. It is true despite Werner Heisenbergs statement: Although natural processes can be described by means of simple laws which can be precisely formulated, these laws, on the other hand, cannot directly be applied to actions in practice. Wherever possible, ofˇcial deˇnitions of IUPAC, ISO and other inter- national organizations have been used, in particular in the \Glossary of Analytical Terms" compiled at the end of the book. However, uniformity could not be achieved in every case. In a few instances, special comments and proposals (characterized as such) have been added. Although progress in the ˇeld of harmonization of nomenclature and deˇnitions has been con- siderable, some things still remain to be done. Some of the contents of various sections have been published previously and are, of course quoted verbatim. In this connection the author is grateful to Klaus Doerffel, Karel Eckschlager, Gunter Ehrlich, Andrzej Parczewski, Karol Florian, Mikulas Matherny, Gunter Marx, Dieter Molch, Eberhard Than, Ludwig Kuchler and others for long-standing collaboration resulting in mutually complementary papers and books. A chapter on statistics was previously published in another book (Accreditation and Quality Assurance Preface VII in Analytical Chemistry edited by Helmut Gunzler at Springer). Gratefully I have adapted essential parts of the translation of Gaida Lapitajs here in this book. Stimulations and ideas have arisen from discussions with William Hor- witz, Duncan Thorburn Burns, Alan Townshend, Koos van Staden, and other members of diverse IUPAC Commissions and Task Groups as well as ofˇ- cers of the Analytical Chemistry Division. From a 20-year membership of various IUPAC bodies I have gained a variety of feedback and experience which has proved to be useful in the writing of this book. In particular, I owe Lloyd Currie and Mattias Otto a great debt of gratitude. They have permitted me to use essential parts of common publications on calibration quoted in Chap. 6. Many of the results stem from work carried out with colleagues, collaborators, postdocs and graduate students. All of them are deserving of my thanks but, due to the large numbers involved, this cannot be done here. Some representative colleagues include Jurgen Einax, Hart- mut Hobert, Werner Schron, Manfred Reichenbacher, Reiner Singer, Diet- rich Wienke, Christoph Fischbacher, Kai-Uwe Jagemann, Michael Wagner, Katrin Venth, Raimund Horn, Gabriela Thiel, Demetrio de la Calle Garc a, and Balint Berente. The author gratefully acknowledges the cooperation with and the support of the editorial staff of Springer. Particularly I have to thank Peter Enders, Birgit Kollmar-Thoni, and { last but not least { John Kirby, the English copy editor, who has essentially improved the English. Many thanks for that. No book is free of errors and this one will be no exception. Therefore, the author would be grateful to readers who point out errors and mistakes and suggest any improvement.
Jena, July 2006 Klaus Danzer Contents
Symbols ...... XIII
Abbreviations and Acronyms ...... XXV
1 Object of Analytical Chemistry ...... 1 1.1 Deˇnition of Analytical Chemistry ...... 1 1.2 Repertoire of Analytical Chemistry ...... 5 References ...... 10
2 The Analytical Process ...... 13 2.1 Principles of Sampling ...... 15 2.2 Sample Preparation ...... 23 2.3 Principles of Analytical Measurement ...... 26 2.4 Analytical Evaluation ...... 31 References ...... 37
3 Signals in Analytical Chemistry ...... 43 3.1 Signals and Information ...... 43 3.2 Analytical Signals ...... 44 3.3 Types and Properties of Analytical Signals ...... 47 3.4 Dimensionality of Analytical Signals and Information ...... 53 3.5 Mathematical Model of Signal Generation...... 60 References ...... 63
4 Statistical Evaluation of Analytical Results ...... 65 4.1 Reliability of Analytical Observations and Measurements . . . 65 4.1.1 Systematic Deviations ...... 67 4.1.2 Random Variations ...... 69 4.2 Uncertainty Concept ...... 75 4.3 Statistical Tests ...... 78 4.3.1 Null Hypotheses ...... 79 4.3.2 Test for Measurement Series ...... 80 4.3.3 Comparison of Standard Deviations ...... 81 4.3.4 Comparison of Measured Values ...... 82 X Contents 4.4 Reliability of Qualitative Analytical Tests ...... 85 4.5 Statistical Quality Control ...... 90 4.5.1 Quality Criteria for Analytical Results ...... 90 4.5.2 Attribute Testing ...... 92 4.5.3 Sequential Analysis ...... 93 4.5.4 Statistical Quality Control ...... 95 References ...... 98
5 Studying In uences and Optimizing Analytical Procedures ...... 101 5.1 Testing the Signiˇcance of In uencing Factors ...... 101 5.1.1 Analysis of Variance (ANOVA) ...... 101 5.1.2 Experimental Design ...... 108 5.2 Optimization of Analytical Procedures ...... 112 5.3 Global Optimization by Natural Design ...... 116 References ...... 120
6 Calibration in Analytical Chemistry ...... 123 6.1 General Fundamentals of Calibration ...... 124 6.1.1 Fundamental and Experimental Calibration ...... 124 6.1.2 The General Three-Dimensional Calibration Model ...... 126 6.1.3 Regression and Calibration ...... 127 6.2 Single Component Calibration ...... 130 6.2.1 Linear Calibration Model ...... 130 6.2.2 Errors in Linear Calibration ...... 134 6.2.3 Weighted Linear Least Squares Estimation (WLS) ...... 137 6.2.4 Linear Least Squares Fitting in Case of Errors in Both Variables ...... 138 6.2.5 Statistical Tests and Validation of Calibration ...... 140 6.2.6 Alternative Calibration Procedures ...... 144 6.2.7 Nonlinear Calibration ...... 151 6.3 Multisignal Calibration ...... 152 6.4 Multicomponent Calibration ...... 155 6.4.1 Classical Multivariate Calibration ...... 157 6.4.2 Inverse Calibration ...... 159 6.4.3 Validation of Multivariate Calibration ...... 162 6.5 Calibration by Artiˇcial Neural Networks ...... 165 References ...... 172
7 Analytical Performance Characteristics ...... 177 7.1 Reliability of Analytical Measurements ...... 178 7.1.1 Precision ...... 178 7.1.2 Precision of Trace Analyses ...... 182 7.1.3 Accuracy and Trueness ...... 183 Contents XI 7.1.4 Remark on the Quantiˇcation of Precision, Accuracy and Trueness ...... 183 7.2 Sensitivity ...... 185 7.3 Selectivity and Speciˇcity ...... 189 7.4 Robustness and Ruggedness ...... 195 7.5 Limit Values ...... 201 7.6 Resolving Power ...... 209 References ...... 212
8 Presentation, Interpretation and Validation of Analytical Results 217 8.1 Presentation of Analytical Results ...... 217 8.2 Factual Interpretation of Analytical Results ...... 219 8.2.1 Presentation of Results Near the Limit of Detection ...... 219 8.2.2 Missing Data ...... 221 8.2.3 Analytical Results in Relation to Fixed Values ...... 224 8.2.4 Interlaboratory Studies ...... 227 8.3 Chemometrical Interpretation of Analytical Data ...... 228 8.3.1 Principles of Data Analysis ...... 229 8.3.2 Cluster Analysis: Recognition of Inherent Data Structures . . 231 8.3.3 Classiˇcation: Modelling of Data Structures ...... 235 8.3.4 Factor Analysis: Causes of Data Structures ...... 239 8.3.5 Exploratory Data Analysis and Display Methods: Visualization of Data Structures ...... 243 8.3.6 Methods of Artiˇcial Intelligence ...... 246 8.4 Analytical Images ...... 250 References ...... 257
9 Assessment of Analytical Information ...... 265 9.1 Quantiˇcation of Information ...... 265 9.2 Information Content of Quantitative Analysis ...... 268 9.3 Multicomponent Analysis ...... 273 9.4 Process and Image Analysis ...... 276 References ...... 280
Glossary of Analytical Terms ...... 283 References ...... 305
Index ...... 309 Symbols
A Matrix of sensitivity factors in (6.66) multicomponent analysis (e.g., ab- sorbance factors) a Regression coefˇcient (calibration coefˇcient): intercept (general and regression in case of errors in both variables) ax Regression coefˇcient a for the esti- (6.9) mation of y from x axw Regression coefˇcient a in case of (6.37) weighted calibration ay Regression coefˇcient a for the esti- (6.10) mation of x from y acc(x), acc(x) Accuracy of an analytical result x (7.10) and of a mean x, respectively (7.11) antilg x =10x , antilogarithm of x b Regression coefˇcient (calibration coefˇcient): slope (general and in case of regression having errors in both variables) bx Regression coefˇcient b for the esti- (6.8) mation of y from x bxw Regression coefˇcient b in case of (6.36) weighted calibration by Regression coefˇcient b for the esti- (6.10) mation of x from y bias(x), bias(y ) Bias (systematic deviation) of an in- (4.1) dividual test value x (or y ) from the (conventional) true value xtrue (or ytrue) XIV Symbols bias(x), bias(y ) Bias of an average test value x (or y ) from the (conventional) true value xtrue (or ytrue)
cnf(x), cnf(y ) Total (two-sided) conˇdence inter- (4.17) vall of a mean cond(A) Condition number of the matrix A (6.80)
cov(pi;pj ) Covariance of the quantities pi and pj CB Conˇdence band (of a calibration (6.26a) line) CR Concordance rate (proportion of (4.53a) correct test results in screening)
dcrit Critical diameter (in microprobe analysis)
dij Distance (in multivariate data) (8.12) df Discriminant function (8.16) dv Discriminant variable
E Information efˇciency
Ein Input energy
Eout Output energy E Residual matrix (in multivariate data analysis)
e, ex , ey Error term (general, in x, and in y , respectively), in concrete terms, ey may be the deviation of an individual value from the mean or the residual of a mathematical model, see next line (ex analogous)
ey = yi y , deviation of an individual measured value from the mean (also dy )
ey = yi y^, deviation of an individual measured value from the estimate of the corresponding mathematical model
F^ Test statistic (estimate) of the F test (4.37) Symbols XV
F1 ˛; 1; 2 Quantile of the F distribution at the level of signiˇcance 1 ˛ and for the degrees of freedom 1 and 2 0 F^max Test statistic (estimate) of Hartley's (4.37 ) test FNR False negative rate (of screening (4.51) tests) FPR False positive rate (of screening (4.50) tests) fdet(:::) Deterministic function of . . . femp(:::) Empirical function of . . .
G^ Test statistic (estimate) of Grubb's (4.36) outlier test
G^max Test statistic (estimate) of Coch- (4.38) ran's test
G˛;n Signiˇcance limits of Grubb's out- lier test for a risk of error ˛ and n individual measurements gaccept(n) Acceptance function in attribute testing greject(n) Rejection function in attribute test- ing gaccept(n;s;˛) Acceptance function in variable test- ing gaccept(n;s;ˇ) Rejection function in variable testing
H 0 a posteriori information entropy H a priori information entropy
H0 Null hypothesis
HA Alternative hypothesis hom(A) Quantity characterizing the homo- (2.9) geneity of an analyte A in a sample
I Information content (9.1)
IAj Speciˇc strength of an in uence fac- tor j on the signal of analyte A
I(p;p0) Divergence measure of information (9.9)
J Information ow (9.40) XVI Symbols L Loading matrix (of principal compo- (6.44) nent analysis and factor analysis) (8.19) LCL Lower control limit (in quality con- trol) LWL Lower warning limit (in quality con- trol)
lx , ly , lz Dimensions (lengths) in x-, y - and z-direction, respectively lb a Binary logarithm of a
M Information amount M(n) Information amount of multicompo- (9.21) nent analysis (of n components)
madfyig Median absolute deviation (around the median), i.e., the median of the differences between an individ- ual measured value and the median of the series: madfyig = medfjyi medfyigjg
medfyig Median of a set of measured values (4.22)
N Number of repeated measurements for the estimation of x by means of a calibration function
Npp Noise amplitude, peak-to-peak noise N ( ; 2) Normal distribution with mean and variance 2 N (t) Time average of noise NPR Negative prediction rate (in screen- (4.55) ing tests) n Number of individual measurements (also in establishing a calibration model), number of objects (in a data set) n(t) Noise component of a signal func- tion in the time domain (i) neti(t), net Net function (propagation function) (6.117) in a neural network (layer i)
(i) Oi(t), O Output of a layer i of a neural net- (6.119) work Symbols XVII P Two-sided level of signiˇcance of sta- tistical tests (P =1 ˛) P One-sided level of signiˇcance of sta- tistical tests (P =1 ˛) P (A) Probability of an event A (ratio of the number of times the event occurs to the total number of trials) P (A) Probability that the analyte A is present in the test sample P (A) Probability of the opposite of the event A (ratio of the number of times the event does not occur to the total number, P (A) + P (A)=1) P (A) Probability that the analyte A is present in the test sample P (AjB) Conditional probability: probability of an event B on the condition that another event A occurs P (T +jA) Probability that the analyte A is present in the test sample if a test result T is positive P Score matrix (of principal compo- (6.44) nent analysis) PB Prediction band (of a calibration (6.26b) line) PPR Positive prediction rate (in screening (4.54) tests) PV Prevalency of screening tests: proba- (4.52) bility that the analyte A is present in a given number of samples p Number of primary samples (in sam- pling); Number of calibration points pi, pj , pij In uence parameters (4.25) prec(x) Precision of an analytical result x (7.8) prd(x), prd(a) Prediction interval of a quantity x (and a, respectively) q Number of subsamples (in sampling) (2.5) q^R Test statistic (estimate) of the David (4.33) test XVIII Symbols
Q Categorial variable characterizing chemical entities (species) being in- vestigated, e.g., elements, isotopes, ions, compounds Q^ Test statistic (estimate) of Dixon's (4.35) outlier test Q(z) Evaluation relationship (of quanti- ties characterizing qualitative prop- erties) in form of a function, atlas, or table Q = f 1(z) Evaluation function of quantities characterizing qualitative properties, viz type of species, Q, in dependence of signal position, z
R Range (R = ymax ymin) R Resolution power R Correlation matrix (6.4)
RC Risk for customers (in quality assur- (4.56b) ance)
RM Redundancy (9.31a)
RM Risk for manufacturers (in quality (4.56a) assurance)
Rt Temporal resolution power (7.57)
Rz Analytical resolution power (7.53) RMSP Root mean standard error of predic- (6.128) tion RR Recovery rate Rfqg Range of a quantity q r Number of measurements at a sub- (2.4) sample
rM Relative redundancy (9.31b)
rxy Correlation coefˇcient between the (6.3) (random) variables x and y
rob(A=B:::; f1 :::) Robustness of a procedure to deter- (7.31) mine an analyte A with regard to dis- turbing components (B;:::) and fac- tors (f1;:::) Symbols XIX rug(A=B:::; f1 :::; u1 :::) Ruggedness of a procedure to deter- (7.33) mine an analyte A with regard to dis- turbing components (B;:::), factors (f1;:::), and unknowns (u1;:::)
S Sensitivity, derivative of the mea- sured quantity (response) y with re- spect to the analytical quantity x
Stotal Total multicomponent sensitivity (7.18) S Covariance matrix (6.5) Sensitivity matrix (7.17)
SAA Sensitivity of the response of the an- (7.12) alyte A with respect to the amount xA of the analyte A (also SA)
SAB Cross sensitivity (partial sensitivity) of the response of the analyte A with respect to the amount xB of the com- ponent B
SAi Cross sensitivity (partial sensitivity) (3.11) of the response of the analyte A with respect to the amount xi of species i (i = B;C;:::N) 0 Sxx, Syy Sum of squared deviations (of x and (6.3 ) y , respectively) 0 Sxy Sum of crossed deviations (of x and (6.3 ) y )
S , S i Sum over a variable index (e.g., at constant i); the dot replaces the in- dex over that the summation is car- ried out S(x) Sensitivity function (6.64) s Estimate of the standard deviation (4.12) (SD of the sample) s2 Estimate of the variance (variance of (4.10) the sample) sN Standard deviation of noise (7.5) spipj , s(pi;pj ) Covariance of the quantities pi (4.27) and pj sy:x Residual standard deviation (of the (6.19) calibration); estimate of the error of the calibration model XX Symbols sel(A;B;:::N) Selectivity of a multicomponent (7.24) analysis with regard to the analytes A;B;:::N snr(y ) Signal-to-noise ratio of an average (7.6) signal intensity y (related to Npp) spec(A=B;:::;N) Speciˇcity of the determination of an (7.26) analyte A with regard to the accom- panying components B;:::N S=N Signal-to-noise ratio of an average (7.1) signal intensity y or a net intensity y net (related to sy ) (S=N)c Critical signal-to-noise ratio (7.51) SS Sum of squares ssr(y ) Sum of squares of residuals (of y ) (6.12)
tW Test statistic (estimate) of the gener- (4.43) alized t test (Welch test) TE Test efˇciency of screening tests (4.53b) TNR True negative rate (of screening (4.49) tests) TPR True positive rate (of screening tests) (4.48) t^ Test statistic (estimate) of Student's (4.41) t test
t1 ˛; Quantile of the t-distribution at the level of signiˇcance 1 ˛ and for degrees of freedom
UCL Upper control limit (in quality con- trol) UWL Upper warning limit (in quality con- trol) u(x) Combined uncertainty of an analyti- (4.31) cal value x
u(x(p1;p2;:::)) Uncertainty of an analytical value x, combined from the uncertainties of the parameters p1;p2;::: u(y ) Combined uncertainty of a measured value y
u(y (p1;p2;:::)) Uncertainty of a measured value y , (4.25) combined from the uncertainties of the parameters p1;p2;::: Symbols XXI
U (x) Extended combined uncertainty (limit) of an analytical value x U (y ) Extended combined uncertainty (4.29) (limit) of a measured value y unc(x) Uncertainty interval of a mean x (4.32) unc(y ) Uncertainty interval of a mean y (4.30)
wyi , wi Weight coefˇcient (in weighted cali- (6.34) bration) x Analytical value: analyte amount, e.g., content, concentration xLD Limit of detection (7.44) xLQ Limit of quantitation (7.48) xQ Analytical value (amount) of a species Q xtest Analytical value of a test sample xtrue (Conventional) true value of a (cer- tiˇed) reference sample x Arithmetic mean of x (of the sample) x^ Estimate of x x(Q) Sample composition (function) x = f (Q) Analytical function in the sample do- main, sample composition function
Y (!) Signal function in the frequency do- main, Fourier transform of y (t) y Measured value, response (e.g., in- tensity of a signal) yc Critical value (7.41) yzi Measured value (e.g. intensity) of a signal at position zi y Arithmetic mean of y (of the sam- ple) y Total mean of several measured se- ries y BL Blank y geom Geometric mean of y (of the sample) y^ Estimate of y XXII Symbols y Outlier-suspected value among the measured values y (t) Signal function in the time domain
ytrue(t) Component of the signal function in the time domain that is considered being true (in uenced by noise) y (z) Signal record (spectrum, chromato- gram, etc.) y = f (x) Calibration function (in the stricter sense, i.e. of quantities characteriz- ing quantitative properties, e.g., sig- nal intensity vs analyte amount) y = f (z) Signal function (measurement func- tion, analytical function in the signal domain)
z Signal position: measuring quantity that depends on a qualitative prop- erty of the measurand. Therefore, analytes may be identiˇed by charac- teristic signal positions. The z-scale may be directly or reciprocally pro- portional to an energy quantity or time zz-scores, standardized analytical val- (8.9) ues z(Q) Signal assignment function (table, atlas etc.) z = f (Q) Calibration function of quantities characterizing qualitative properties, viz signal position as a function of the type of species
˛ Risk of the error of the ˇrst kind (two-sided) ˛ Regression coefˇcient (intercept) (4.2)
˛i Coefˇcient characterizing in uences (5.3) ˛ Risk of the error of the ˇrst kind (one-sided)
ˇ Risk of the error of the second kind (two-sided) Symbols XXIII ˇ Regression coefˇcient (slope) (4.3)
ˇj Coefˇcient characterizing in uences (5.4) ˇ Risk of the error of the second kind (one-sided)